CIS 5590.004. Artificial General Intelligence

Concept and Learning

1. Models of Concept

Psychological models:

Classical view: necessary and sufficient condition
Prototype theory, exemplar theory, "theory" theory
Concept hierarchy and taxonomies, basic-level concepts

Formal models:

Definition: set, category, class, frame
Graph: Semantic networks, Conceptual graphs, Knowledge graphs
Space: Conceptual space, Semantic space, Vector space

Advanced topics:

2. Conceptual learning

In current machine learning:

Learning conceptual relations: supervised learning, end-to-end training of a universal function approximator
Learning classifications: unsupervised learning, clustering, associative memory

3. Set-Theoretic Inference Rules in NAL

Increasing the intelligence of NARS: richer experience, more efficient resource uses

Steps in extending the model: (1) Narsese, (2) IL, (3) NAL, (4) NARS

NAL-2, NAL-3, NAL-4: copulas and compound terms inspired by set theory

Concept model: extension and intension, graded and dynamic, empirical and analytical
Concept graph and its boundary
Copulas vs. conceptual relations
Compound terms and concept composition
Concept meaning: general, current, literal
Learning as reasoning, multi-strategies

Reading

Non-Axiomatic Logic: A Model of Intelligent Reasoning, 2nd Edition, Chapters 6-8